MOSFET Characteristics at Cryogenic Temperature

Juan M. Avalos Electrical Engineering and Computer Science Department

Christian Brothers University

Memphis, TN USA

javalos@cbu.edu

Leon Tolbert

Electrical Engineering and Computer Science Department

University of Tennessee

Knoxville, TN USA

tolbert@utk.edu

Abstract— Efficiency is a large concern in power electronics but

size and weight must be taken into consideration as well. Power

devices in need of thermal sinks for appropriate performance will

affect both size and weight. MOSFET devices are the main

components in the power electronics world because of their

comparatively promising characteristics in power loss and

switching performance. The objective of this research is to obtain

results as to how the MOSFET device will perform in various

temperatures that go as low as liquid nitrogen temperatures.

More than one device was tested and compared in order to get the

best performing. It is believed Cryogenic operation of MOSFET

devices results in improved performance and efficiency. The

switching performance of the devices will be analyzed using the

Double Pulse Test (DPT). Ultimately, the objective of this research

is to design a DC-AC inverter for Boeing & NASA aircraft using

the best performing MOSFET device.

Index Terms—Double Pulse Test, Power Inverter, Cryogenic

Temperature

I. INTRODUCTION

It is essential to understand the structure of MOSFET

devices and the Double Pulse Test process that is used to

analyze them. There are four main characteristics that are being

observed for each device. Each defines the device’s efficiency

and performance. On-state resistance, switching loss, body

diode, and breakdown voltage are the four characteristics that

will be affected by the change in temperature and determine the

device performance gains. Different techniques are used to

obtain each one and will be discussed further on.

The purpose of the Double Pulse Test is to obtain data

reflecting the switching characterization of power devices in a

simplified printed circuit board (PCB). This board allows to

evaluate their switching behaviors as if in an actual power

converter. The Double Pulse Test consists of manually

triggering two pulse as an input to the gate of the device. Each

pulse has its purpose and effect to the device. The length of the

first pulse, in time, builds up a desired current through the

inductor load. The switching transient consists of the turn off

and turn on of the device. The falling edge of the first pulse

corresponds to the turn off transient while the rising edge of the

second pulse corresponds to the turn off.

The components required for this set up are the following:

Device Under Test (DUT)

Inductive Load

Auxiliary DC Power Supply

Waveform Generator

DC bus

Gate drive/driver board

Oscilloscope

Lab procedures for testing devices have to be taken

into consideration for accurate results. Initially, probe

compensation and de-skew correction are necessary for

voltage-current alignment. The yellow scope must compensate

for the input at the gate of the MOSFET while the blue scope

compensates for the drain source voltage. As it is usually

assumed, the channel 1 with yellow waveforms is used as gate-

source voltage measurement, the channel 2 with blue

waveforms is used as drain-source voltage measurement, the

channel 3 with purple waveforms is used as drain-current

measurement, and the channel 4 with green waveforms is used

as load-current measurement. The purple current probe is de-

skewed and the propagation delay for all are minimized as

much as possible. The probes have different purposes such as

the probes being used in channels 1 and 2 for voltage

measurement. The difference in those two channels are the

voltage levels: 1 for ≤ 600V; 2 for 600V ~ 1200V. Different

probes will have different parasitic, which will cause the error

of the measurement, in order to avoid that, the de-skew process

was conducted. Meanwhile, when the 600V-level test was

conducted, same type of probes was used for both channel 1 and

2, the de-skew process was not necessary in this case.

Before any testing is done to the devices, the

circuit board must have exposed, conductive parts covered.

Icing at lower temperatures can cause problems such as shorts

in the circuit leading to damaging equipment. The lowest

temperature, for the chamber, allowed is 93 K (-180°C) and the

highest temperature tested is 23 K (room temperature). Once

the board is ready, it is placed inside the chamber even for room

temperature testing. The room temperature testing is done after

the chamber has been heated up for moisture dissipation. This

procedure mitigates humidity in the inner chamber and reduces

the possibility of short-circuiting due to icing at cryogenic

temperature.

Fig. 1. Icing at cryogenic temperature

The circuit board consists of several parts having

voltage supplied to it. The gate driver has a 25 V DC supplied

to it and the 250 V DC auxiliary power is the main supply for

the circuit. The device rated at 500 V is supplied with 250 V

from the auxiliary power. Other devices rated at higher

voltages such as 800, 900 and 1200 V are supplied with up to

500 V from the bus voltage. The last input for the test would be

using the Tektronix program to generate a double pulse gate

signal. For more accurate results, the Auxiliary power supply

for gate the drive does not function properly at cryogenic

temperature; therefore, placed outside the chamber. The first

pulse is varied in time which increases the current induced in

the device. This changes the switching characteristics, which

is what is being analyzed. The falling edge of the first pulse is

the turn off region while the rising edge of the second pulse is

the turn on. This switching transient region is analyzed for the

device’s switching characteristics.

Scaling and positioning may have to be adjusted on the

oscilloscope when showing results. Every value of the transient

switching time must be shown on the display. Overshooting

and ringing are the main causes of displaying problems, which

add difficulty of MATLAB calculations of the switching

characteristics: energy loss, switching time, delay time, etc.

After adjusting, the double pulses are inputted again and

repeated until the right scaling and positioning is acquired. The

saving process includes the time periods and values of the

switching transient, which are then implemented in a MATLAB

code. Values for time delays of the drain source voltage, gate

voltage, and current are obtained using code.

The four characteristics are broken down and explained

in the following sections. Each concluding with results from

experimentation and graphs using the data obtained. The

experiment consists of more than one device from different

manufacturers: Microsemi, Infineon, IXYS

II. ON-STATE RESISTANCE

The University of Birmingham tested the

SPP20N60C3 power MOSFET for a 500 V converter and

obtained its characteristics. These higher voltage devices result

in a larger reduction of on-state resistance at cryogenic

temperatures [1]. The on-state resistance consists of more than

one component; the drift region resistance of MOSFET devices

dominate the on-state resistance when devices are rated above

150 V [4]. This is due to its temperature sensitivity [1]

Components of the MOSFET resistance [2]:

RDS(ON) = Rsource + Rch + RA + RJ + RD + Rsub

Source diffusion resistance

channel resistance

accumulation resistance

JFET component-resistance

drift region

substrate

The following shows the basic components

of a power MOSFET which include major contributors to the

device’s resistance [4].

Figure 2: MOSFET components

The Lab procedure consists of using a high power

curve tracer which plots the devices characteristic curve. The

plot is labeled as the following: (VDS) drain-source voltage on

horizontal axis and (ID) drain current on the vertical axis. (VGS)

Gate-source voltage is a constant input for each individual

curve, normalized throughout 10 V to 20 V.

Figure 3: On-state Resistance & Body Diode set-up

The resulting curves will allow to find the drain-source

voltage that corresponds to the rated drain current

recommended in the device’s datasheet at various temperatures.

The two values can then be used to calculate the on-state

resistance for each individual device at a certain temperature.

The experiment began at room temperature and decreased to -

180 degrees Celsius in intervals of 50 degrees, resulting in 6

plots for each device.

Figure 4: Curve Tracer Results (VDS vs. ID)

In closing, the RDS-ON significantly reduces as

temperature drops. Due to the improvement of carrier mobility

in the drift region, conduction loss during operation is reduced

[1].

Plot of on-state resistance using curve tracer results:

Figure 5: On-state Resistance Results

III. SWITCHING LOSSES

MOSFET devices have the ability to switch from on

to off in a very small amount of time. Devices differ and behave

differently; therefore, some devices are more efficient than

others in terms of switching.

Load

Inductor

Upper

Diode

Lower

Switch

– +

DC Link

Capacitor

Gate

Driver

DC

source

Signal

Generator

Signal

Isolator/

Logic

AUX

Power

Supply

Scope

Power

Supply

Cryo. Chamber

Figure 6: Set-up for Switching Testing

As earlier stated, the procedure used is called the

double pulse test and allows for the observation of delays and

propagations in the switching transient. The double pulse test

consists of two pulses that are inputted into the gate; the first

falling edge of the first pulse is the turn off transient while the

rising edge of the second pulse is the turn on transient. Before

testing the devices, the probes being used must be well

calibrated and the V-I alignment must be done. Within the

probes being used, the voltage and current probes are ones that

have different characteristic propagation delay and are de-skew

corrected for accuracy enhancement. This process is done for

each individual test.

MOSFET devices have internal parasitic that effect

the behavior in the switching transient. Parasitic inductances

will cause ringing and overshooting; parasitic capacitance

determines a time constant for the circuit and increases delay.

The results in the oscilloscope show the following: VGS, VDS, ID

and inductor current.

The ringing due to parasitics is shown in a turn on transient

response [5]:

Figure 7: DPT Switching Characteristics

These results lead to the calculation of TON_DELAY, TON,

TOFF_DELAY, TOFF, EOFF and EON via MATLAB code. The code

calculates average values over a certain period of time and

value of energy loss. The delay times relate to the amount of

time difference between the current and voltage at an off or on

transient. On and off time corresponds to just the current. It is

the amount of time it takes the current to change within 10% to

90% of its value.

The MOSFET switching performance improves at

lower temperatures, resulting in changes of the calculated

values. The energy loss and the TON for the on period is

significantly reduced while the EOFF and TOFF are only slightly

reduced. The delay times both roughly remained the same

value as temperature decreased. The turning on switching

transient shows the most improvement in the device’s testing at

lower temperatures.

As the device is in the off mode/cut off region, the

reverse drain current is now passing through the body diode

which shorts the source and drain [2]. This creates a high

current path for the device, rather than the drain-source

conducting channel [2]. Reverse recovery characteristics of

the body diode play a significant role in the switching

characteristics of the device. During switching transients, the

reverse recovery is divided into time periods as seen below.

The reverse recovery time is an important factor of losses and

defines the time it takes for drain current to reach a constant

value of zero amperes.

Figure 8: Reverse Recovery behavior

In order to decrease switching and conduction losses,

the input from oscilloscope now includes a dead time. Dead

time is an input that is between two other pulses in which the

inductor current is taking the body diode path.

IV. BREAKDOWN VOLTAGE & BODY DIODE

The power MOSFETs can be operating in three

different stages: cut-off, saturation, and breakdown region. The

breakdown region limits the drain-source voltage amplitude of

the device and is located to the far right of the MOSFET

characteristic curve. The breakdown region varies while it

depends on the gate-source input and temperature. When the

device receives too much voltage across the drain-source

terminals, it reaches the point where the device cannot resist

current and breaks down. This allows a large amount of current

to flow through the device and drastically increases. Due to a

correlation with temperature, on-state resistance shows a non-

linear relationship with the breakdown voltage when plotted.

Figure 9: Breakdown voltage testing schematic

The breakdown voltage configuration allows the

measurement of current when the device is reaching its rated

voltage. Initially, the current is so small it is considered to be

zero and 1 mA when at its breakdown. The devices under

testing had a smaller breakdown voltage then the voltage they

were rated at, when operating at lower temperatures.

Figure 10: Breakdown Voltage Results

The configuration shows the device is set up in series

with a large resistance of 100 kOhms that will limit the current.

While the gate and source are short circuited, voltage is applied

to see how much is needed in order for the ammeter to read

current. The breakdown voltage is set as to when the leakage

current reaches 1 mA.

The diode forward voltage is the amount of voltage

drop across the diode when the source has a specified value of

current. The usual current used is the rated current from the

device’s datasheet.

The following figure shows the configuration when using the

curve tracer for on-state resistance and body diode testing:

Figure 11: Schematic for RDS-ON & Body Diode

V. 77K LIQUID NITROGEN POURING TEST

The lowest temperature used on the devices under

testing was the liquid nitrogen temperature of about 77 K.

Having a minimum temperature operation of about 93 K, the

cooling chamber Delta 9023 model was not functional for the

lowest temperature. Using a Styrofoam container and a larger

plastic container, each device was directly inserted into the

liquid nitrogen while under testing. The previous schematics for

the tests still apply for the 77 K testing; therefore, same

procedures apply.

The time for this is quite limited because the liquid

nitrogen will last about 30 minutes in its liquid state and will

evaporate. This evaporation will cause the temperature to start

to increase again. In order to keep the temperature constant at

around 77 K, liquid nitrogen is constantly being poured in small

amounts into the Styrofoam container. The larger plastic

container is uses as an outer shell in case of leaks or spills.

The image below shows the set up for the lowest temperature

testing:

–+

A

D

G

S

High voltage

DC source

Current limiting resistor

Ampere meter

Device

under

test

(DUT)

Cryo. chamber

450

500

550

600

650

700

750

70 120 170 220 270 320

Bre

ak

do

wn

Vo

lta

ge

(V)

Temperature(K)

VI. CONCLUSION

This research demonstrates a way to reduce the losses

in power MOSFETs. This allows for more efficient power

devices with many advantages when operating at cryogenic

temperature. MOSFET devices have switching characteristics

that will have losses at high frequencies. Under low

temperatures, the devices have slightly faster switching with

less ringing, which is caused by internal parasitic as earlier

stated. On-state resistance drastically decreased due to the

increase of carrier mobilities in the MOSFET’s main resistive

components. These factors allow the efficiency of the device

to increase and enhance its performance at cryogenic

temperatures.

ACKNOWLEDGMENTS

This work was supported primarily by the Engineering

Research Center Program of the National Science Foundation

and the Department of Energy under NSF Award Number

EEC-1041877 and the CURENT Industry Partnership Program.

The help and guidance from Dr. Leon Tolbert and Jingyi Tang,

from the University of Tennessee Knoxville, was greatly

appreciated. All results are remained to be owned and explained

by Boeing Company and NASA supervision.

REFERENCES

[1] C. Jia, “Experimental Investigation of Semiconductor Losses in

Cryogenic DC-DC Converters,” Dept. Elect. Eng., Univ. of

Birmingham, 2008.

[2] V. Barkhordarian. Power MOSFET Basics. [PDF file]. Available:

http://www.infineon.com/dgdl/mosfet.pdf?fileId=5546d4625336

00a4015357444e913f4f

[3] A. Sattar. Power MOSFET Basics. [PDF file]. Available:

http://www.ixys.com/Documents/AppNotes/IXAN0061.pdf

[4] J. Dodge. (2006, March 2). Power MOSFET Tutorial. [PDF file].

Available:http://www.microsemi.com/document-

portal/doc_view/14692-mosfet-tutorial

[5] A. S. J. H. E. S. a. R. P. A. Caiafa, "IGBT operation at cryogenic

temperatures: non-punch-through and punch-through

comparison," in Power Electronics Specialists Conference, 2004.

[6] G. D. H. L. U. L. H. S. J. S. A.E. Schlogl, "Properties of CoolMOS

between 420 K and 80 K-the ideal device for cryogenic

applications," in Power Semiconductor Devices and ICs,

Toronto, Ont., 1999.

[7] A. ,. K.K.Leong, "Power MOSFET operation at cryogenic

temperatures: Comparison between HEXFET®, MDMeshTM

and CoolMOSTM," in International Symposium on Power

Semiconductor Devices & IC's, Hiroshima, 2010.

[8] C. C. D. T. D. P. Thibaut Chailloux, "Characterization and

comparison of 1.2kV SiC power," Materials Science Forum, pp.

814-817, 2015.

[9] T. K. A. R. Y. P. R. L. D. Z. R. W. X. W. Y. J. S. G. K. H. W. P. J.

B. a. A. C. Paul Gemin, "Architecture, Voltage and Components

for a Turboelectric Distributed Propulsion Electric Grid (AVC-

TeDP)," NASA, 2015., 1989